fix: strictly output [START], [STEP], [END] to stdout for phase 2 validator parsing

fix: restore UI, add /tasks /grade, clamp scores to (0.05, 0.95)

Massive Rewrite: Rebuilt Phase 2 standalone validation constraints perfectly mapped to 0.05-0.95

Fix inference.py output format to strictly match Step/End validation criteria per task

hf done

hf

good to go

feat: implement SQL agent environment with OpenEnv integration, SQLite management, and automated task grading